The AI labs are running out of runway, and it's starting to show in their product decisions.

The Summary

  • Both Anthropic and OpenAI are facing a "monetization cliff" as hundreds of billions in capital investment demands returns before the bubble pops.
  • AI agents (Claude Code, Cowork, OpenAI's Codex) are forcing these companies to radically rethink resource allocation and product strategy.
  • The pressure to go public is changing behavior: which products live or die, what restrictions get imposed on customers, and how much cash they'll burn chasing the next breakthrough.

The Signal

The AI industry is hitting its "show me the money" moment, and the cracks are visible. Anthropic and OpenAI, the poster children of foundation model development, are both scrambling to turn chatbot popularity into actual business models before investor patience runs out. The math is brutal: billions spent on compute, data centers, and chips, with revenue models that still look more like subsidized demos than sustainable businesses.

What's forcing the shift isn't abstract market pressure. It's agents. Products like Claude Code and Cowork represent a fundamental pivot from "better autocomplete" to "software that does jobs." This matters because agents burn resources differently. They're not one-off queries. They're persistent, they iterate, they rack up API calls. The cost structure changes completely when your product is running complex multi-step workflows instead of answering single questions.

This resource crunch is already changing how these companies operate. Products are getting killed faster. Customer restrictions are tightening. The era of "scale first, monetize later" is ending because the capital markets are demanding a path to profitability, not just a path to AGI. The Verge's Hayden Field is tracking something important here: these aren't just business decisions, they're existential ones. If you can't show how agents translate to revenue before the next funding round, you might not get a next funding round.

The shift also signals a deeper market maturation. The AI hype cycle sold AGI as the endgame. But the actual money will come from narrower, specific-use automation. Agents that write code, agents that handle customer service, agents that do compliance work. Boring stuff that companies will actually pay for at scale.

The Implication

Watch how OpenAI and Anthropic price their agent offerings in the next six months. That will tell you whether they've found a real business model or are still stalling. For builders in this space, the message is clear: if the foundation model companies can't make agents profitable, your venture-backed agent startup probably can't either. Focus on narrow, high-value use cases where customers will pay real money, not just pilot budgets. The party's ending. Build something that generates revenue, not just engagement.

Source: The Verge AI